Describe a rack focus as a change in which depth plane is sharp: name the first target, the second target and the moment attention transfers. Also say what should stay fixed. A request to make a distant face sharp cannot guarantee recovery of identity detail that is absent from the reference.
Runway's camera terminology guide defines rack focus as a focus change between subjects or objects. It is a different instruction from moving the camera closer.
Choose two readable targets
Use concrete visible objects rather than “cinematic focus.” For example, a ring box occupies the near foreground, while a partner stands farther back. The viewer needs enough composition and contrast to understand both targets before and after the transition.
Avoid combining the first test with a camera orbit, a walk, a major lighting change and a focus pull. When several things change at once, you cannot tell whether the focus instruction worked. Add further movement only after the simpler shot has a usable result.
Write down the intended end state: the partner's face is the sharp target, the ring box is softer, and the composition remains recognizable.
Check what the reference can actually show
Inspect the second target at its available source size. Is the face large enough to identify? Are the eyes and mouth visible? Is it obscured or already strongly blurred?
A model may invent plausible detail when asked to reveal a blurry area. Treat that detail as generated, not recovered evidence. If identity matters, use an authorized clearer reference where the workflow supports it, or change the composition so that the shot does not depend on revealing missing detail.
As a related still-image boundary, Adobe's Camera Raw Lens Blur documentation says its Focus and Blur refinement options require an original image that is not blurry. This is not a promise that Camera Raw creates video focus pulls or restores an unknown face.
An original fictional shot brief
For a proposal film, the planned brief is:
A fixed camera frames an open ring box in the near foreground and the partner's visible face farther back. Begin with the ring in focus. As the partner notices it, focus transfers to their face. Hold that focus long enough to read the reaction. Keep the camera position and both subjects' placement steady.
This is an invented planning example, not a claim that a model has executed the shot. Its purpose is to make the requested change observable. No exact lens setting is needed unless the production workflow can actually control or reproduce it.
Review the transition, not only its endpoints
Check the generated clip with five questions:
- Is the first sharp target clearly identifiable?
- Does emphasis transfer to the intended second target?
- Does the second face keep the approved identity?
- Do geometry and placement remain coherent during the transition?
- Is there enough time at the end to understand the reaction?
A soft background becoming sharper is insufficient if the face changes. A camera push that enlarges the person is also a different result, even if it attracts attention.
Revise the smallest failed instruction
If focus does not transfer, simplify the scene or make the target separation clearer. If identity fails, improve the supported reference information or use another shot. If the transition produces unstable edges, a cut between two accepted shots may communicate the moment more reliably.
Plan the focus change and review criteria in Imagild Film Studio. Accept a generated shot by its visible behavior rather than by the camera term in its prompt.